Relation Extraction for Chinese Clinical Records Using Multi-View Graph Learning

Information Extraction from the Text Data on Traditional Chinese Medicine: A Review on Tasks, Challenges, and Methods from 2010 to 2021

“…In the pretraining process, the pretraining BERT model was more accurate in learning contextual features [61]. Taking into account that there were many types of complex relation existing between many heterogeneous medical entities, Ruan et al [21] proposed a multiview graph representation learning method that adopted the heterogeneous graph convolutional network (GCN) and attention mechanism for the extraction of complex relationships among multitype entities. ey first constructed a heterogeneous entity graph from the co-occurring relations and used domain knowledge to enhance the semantic information of the above graph to infer the labels of all the candidate relations simultaneously by employing node representation learning and classification.…”

“…In this review, some studies reported that the application or combination of BERT could significantly improve the result of entity recognition or relation extraction [18,58,59,61]. In the last decade, the proposed deep learning models for IE tasks include BERT-convolutional neural network (CNN) [18], convolutional neural network with segment attention mechanism (SEGATT-CNN) [63], K-nearest neighbor (KNN) [53], long short-term memory (LSTM) [52,53], bidirectional long short-term memory (BiLSTM) [17], structural BiLSTM [31], LSTM-CRF [54,58], BiLSTM-CRF [22,28,55,58,62,64], BERT-BiLSTM-CRF [59,61,66], graph neural networks [21], and a nested NER model based on LSTM-CRF [29]. Among the above-mentioned models, the "BiLSTM-CRF" and "BERT-BiLSTM-CRF" have become popular deep learning models because of their good extraction performance: the BiLSTM model can capture more context information than the LSTM model.…”

“…With this background, more approaches were explored to extract information from different types of TCM text data, and IE from TCM texts has shown encouraging improvements accordingly [17,18]. Although previous research has summarized some IE work in TCM [19][20][21], the new advanced technologies and emerging methods need to be further summarized and synthesized, for example, improved deep learning approaches and more types of extracted information [22,23]. In this study, we searched four literature databases for articles published from 2010 to 2021 that focused on the use of NLP methods to extract information from unstructured TCM text data.…”

Systematic Literature Review of Information Extraction From Textual Data: Recent Methods, Applications, Trends, and Challenges

“…For example, NER can be used to identify and extract named entities and types of entities from a text body. RE is used to extract relational information between entities, whereas EE is used to identify and extract event occurrences and their arguments from text [25], [63], [67], [72].…”

“…Thus, each NER extraction model is limited by the language it is trained to extract [6], [11]. A few examples include NER in Portuguese texts [46], [91], Chinese texts [72], [76], [89], [127], [128], [129], Indonesian texts [65], [78], [87], [119], [130], [143], Malay texts [92], and Arabic texts [21]. This demonstrates that language is a common barrier to extracting useful information from text documents.…”